Anda di halaman 1dari 12

Mobile Netw Appl

DOI 10.1007/s11036-016-0711-y

Optimal Energy Harvesting-based Weighed Cooperative


Spectrum Sensing in Cognitive Radio Network
Xin Liu1,2 Kunqi Chen2 Junhua Yan2 Zhenyu Na3

Springer Science+Business Media New York 2016

Abstract In cognitive radio (CR) network, to improve


spectrum sensing performance to primary user (PU) and
decrease energy wastage of secondary user (SU) in cooperative spectrum sensing, an energy harvesting-based weighed
cooperative spectrum sensing is proposed in this paper. The
SU harvests the radio frequency (RF) energy of the PU signal and then converts the RF energy into the electric energy
to supply the power used for energy detection and cooperation. The time switching model and power splitting model
are developed to realize the notion. In the time switching
model, the SU performs either spectrum sensing or energy
harvesting at any time, while in the power splitting model,
the received PU signal is split into two signal streams,
one for spectrum sensing and the other one for energy
harvesting. A joint optimization problem is formulated to
maximize the spectrum access probability of the SU by
jointly optimizing sensing time, number of cooperative SUs
and splitting factor. The simulation results have shown that
compared to the traditional cooperative spectrum sensing,
the proposed energy harvesting-based weighed cooperative

 Xin Liu

liuxinstar1984@nuaa.edu.cn
1

School of Information and Communication Engineering,


Dalian University of Technology, Dalian 116024, China

College of Astronautics, Nanjing University of Aeronautics


and Astronautics, Nanjing 210016, China

School of Information Science and Technology, Dalian


Maritime University, Dalian 116026, China

spectrum sensing can decrease the energy wastage obviously while guaranteeing the maximum spectrum access
probability.
Keywords Cognitive radio Energy harvesting
Cooperative spectrum sensing Spectrum access

1 Introduction
In cognitive radio (CR), the primary user (PU) occupies
the spectrum resources licensed by the government with the
preferential access opportunity [1]. In order to improve the
current spectrum utilization, CR allows the secondary user
(SU) to dynamically access the idle spectrum, which is temporarily unused by the PU, providing that the SU will not
disturb the normal communication of the PU [2]. The SU
detects whether the PU exists in the frqeuqncy spectrum
depending on spectrum sensing, and only the absence of the
PU is detected, the SU can access the PU spectrum [3].
As the CR and the PU are two different kinds of
communication systems, energy detection, which can be
implemented easily and facilely without acquiring any prior
information from the PU signal, has been widely used in
CR as an effective spectrum sensing method [4]. However,
it has shown lower correct detection performance if the PU
is shadowed or in severe fading. Hence, cooperative spectrum sensing is proposed to overcome this problem, which
may obtain a sensing diversity gain through combining the
sensing results of multiple SUs [5, 6]. The most important problem is that cooperative spectrum sensing has to
deplete some electric power such as local sensing power,

Mobile Netw Appl

cooperative power, et al, thus yielding the decrease of the


transmission power. [7, 8] investigate an energy-efficient
cooperative spectrum sensing to maximize the spectrum
sensing performance while using less spectrum sensing
power. However, since the outside energy cannot be utilized
such as the energy of the PU signal, the cooperative spectrum sensing still needs to consume some energy that is
originally used for data transmission.
Energy harvesting has recently been proposed to supply the electric power for a wireless communication system
instead of a fixed power supply, through collecting the radio
frequency (RF) energy from environmental signal sources
[9]. The energy-harvesting circuit is designed to convert the
received RF energy to direct current (DC) power by using
semiconductor-based rectifying elements [10]. In [11], an
energy-harvesting CR network is investigated, where the RF
energy of the PU signal is harvested by the SU, however,
the energy harvesting and the spectrum sensing are implemented independently, thus the energy-harvesting efficiency
is very low, and the harvested energy cannot supply cooperative spectrum sensing in time. Since the sensed PU signal
that carries energy can be used for energy harvesting at the
same time, simultaneous spectrum sensing and energy harvesting will become a profitable research, which may save
the transmission energy through harvesting the RF energy
of the PU signal to supply spectrum sensing.
We attempt to pursue a combination study on cooperative
spectrum sensing and energy harvesting for CR with such
a dual usage. The contributions of the paper are listed as
follows:
(1) We have proposed two combination models: time
switching model and power splitting model. In the
time switching model, the SU acts as a detector or
a harvester separately, the SU senses the PU within
some time and then harvests the energy of the PU signal within the remainder time. In the power splitting
model, the SU acts as a detector and a harvester simultaneously, the SU uses part of the power of the PU
signal to sense the PU and the remainder power to
harvest energy.
(2) Based on the proposed combination model, a joint
optimization problem of sensing time, number of
cooperative SUs and splitting factor is proposed to
maximize the spectrum access probability of the SU.
Through optimization, the SU can obtain enough
spectrum access, while guaranteeing better spectrum
sensing performance and lower energy wastage.
(3) Using the alternating direction optimization (ADO),
we have proposed a joint optimization algorithm to
solve the proposed optimization problem.

2 Weighed cooperative spectrum sensing


2.1 Energy detection
In CR network, each SU is difficult to obtain any prior information of the PU signal, hence, energy detection, which
can be performed without needing any information of the
detected signal, is used to sense the PU as an effective
spectrum sensing method. In energy detection, the detected
signal yi received by the SU i is given by a binary hypothesis
problem as follows [12]

yi (t) =

n(t), H0
, t = 1, 2, ..., M
hi (t)s(t) + n(t), H1

(1)

where H0 and H1 denote the absence and presence of the


PU, respectively, s(t) is the PU signal with the power ps ,
n(t) is the Gaussian noise with the power n2 , hi (t) is the
channel gain from the PU to the SU i, M is the number of the
signal samples. We also suppose that P (H0 ) and P (H1 ) are
the absent and present probabilities of the PU, respectively.
If the sampling frequency is fs and the sensing time is , M
is given as follows
M = fs

(2)

From Eq. 1, the energy statistic of yi is given as follows


M
1 
yi (t)2
(yi ) =
M

(3)

t=1

Since yi (1), yi (2), ..., yi (M) are independently and identically distributed, from Eq. 3, with a large M, (yi ) obeys
the Gaussian distribution as follows

4
N 2 , n , H0
n M


(yi )
N (1 + i ) 2 , (1+i )2 n4 , H1
n
M

(4)

where i = h2i ps /n2 is the sensing signal to noise ratio


(SNR). By comparing the energy statistic of the PU signal to
a properly set decision threshold, energy detection declares
the presence of the PU with the energy statistic lying
above the threshold, while deciding the absence of the PU
with the energy statistic lying below the threshold. Hence,

Mobile Netw Appl

we compare (yi ) to a threshold i , and then the probabilities of false alarm and detection are respectively given as
follows

 

P f = Pr ((yi ) i |H0 ) = Q i2 1
fs
i
 
 n

d
i
P = Pr ((yi ) i |H1 ) = Q

1
fs
i
2 ( +1)
n

(5)
where the function Q(x) =

1
2

+
x


exp z2 /2 dz.

2.2 Weighed cooperative spectrum sensing


However, if the PU is in deep fading or shadowed, the performance of energy detection can be decreased greatly, and
one of the ways to improve spectrum sensing reliability is
through the cooperative spectrum sensing. In cooperative
spectrum sensing, each SU senses the PU to obtain local
sensing information and the final decision on the presence
of the PU is made by performing energy detection using the
combined energy statistic of multiple SUs. The performance
of cooperative spectrum sensing is improved by achieving
a sensing diversity gain provided by the different sensing
paths from multiple SUs, and even though one SU has failed
to detect the presence of PU, other SUs may also help to
detect the PU accurately, as shown in Fig. 1.
The frame structure of CR network with cooperative
spectrum sensing is shown in Fig. 2, which is divided into
local sensing slot, cooperative slot and transmission slot. In
the local sensing slot, each SU senses to obtain the energy
statistic of the PU independently; then in the cooperative
slot, all the SUs exchange their local sensing information

Fig. 2 Frame structure of CR network

through a public control channel, a combined energy statistic is obtained through combining all the exchanged local
sensing information of SUs, and the combined energy statistic is compared to a preset threshold to get a final decision
on the presence of the PU; finally, in the transmission slot,
if the absence of the PU is determined, the SU will transmit
data, otherwise, the SU must wait to re-sense the PU in the
next frame.
Supposing that there are k SUs to perform cooperative
spectrum sensing, from Eq. 3, the combined energy statistic
at each SU is given as follows
(y) =

k


(6)

i (yi )

i=1

where i for
i = 1, 2, ..., k are the combined weights
that satisfy ki=1 i2 = 1. (y) is compared to a global
detection threshold . By substituting Eq. 6 into Eq. 5, the
cooperative probabilities of false alarm and detection are
respectively given as follows

 

k

fs

Q = Q n2 i=1 i


(7)
2 k ( +1)

/
i
i

n i=1

f
Qd = Q
s

k
2 ( +1)2
i=1 i

In CR network, as the distance between the PU and the


SU is often much larger than that between the SUs, we may
assume that i for i = 1, 2, ..., k are not very different.
Hence, from (7), Qf is related with Qd approximately as
follows


 k
f
1
d
i i
(8)
Q = Q Q (Q )(1 + ) + fs
i=1


where the average sensing SNR is = ki=1 i /k. Since
Q(x)
is a monotonously decreasing function, we can maximize ki=1 i i to minimize Qf as follows


k
k
k
k
2
2
i i
i
i =
i2
(9)
i=1

Fig. 1 Cooperative spectrum sensing model

i=1

i=1

i=1

where
 the maximum is achieved only when i =
k
2
i /
i=1 i for i = 1, 2, ..., k. By substituting (9) into
(8), the minimum false alarm probability is given as follows



k
f
1
d
Qmin = Q Q (Q )( + 1) + fs
i2
(10)
i=1

Mobile Netw Appl

3.2 Energy harvesting-based cooperative spectrum


sensing model

Fig. 3 Energy-harvesting SU model

3 System model
3.1 Energy harvesting-based SU
In this paper, we consider an energy-harvesting SU that
harvests the RF energy of the PU signal and converts
the energy to the electric power to supply the cooperative spectrum sensing, as shown in Fig. 3. The SU
stores the arriving energy from the PU in a rechargeable battery through deploying an energy-harvesting circuit, and the harvested energy is used to supply the consumed energy of cooperative spectrum sensing including
local sensing power and cooperative exchange power. The
energy-harvesting circuit is comprised of band-pass filter,
rectifying circuit and low-pass filter and designed to convert the ambient RF energy to the DC power. When the
PU is really present in the spectrum, the SU may harvest
the RF energy of the received PU signal in the local sensing slot.
The energy-harvesting circuit is shown in Fig. 4, wherein
the SU firstly passes the received PU signal through a
band-pass filter to suppress the out-of-band interference,
then converts the PU signal to the DC signal through
a rectifying circuit, and finally outputs the DC voltage
after filtering out the fundamental and harmonic signals from the DC signal through a low-pass filter. However, some of the signal energy may be reradiated to
the outside environment in the energy-harvesting process,
and we assume that 0 < < 1 is the electromagnetism-to-electricity conversion efficiency that is determined by the element characters of the energy-harvesting
circuit.

Since the SU needs the power to perform cooperative spectrum sensing, in this paper, we attempt to pursue a combined
spectrum sensing and energy harvesting for CR in order to
make the SU use the harvested energy to supply cooperative
spectrum sensing.
In energy harvesting-based cooperative spectrum sensing, the converted electric power from harvested energy
is used to supply the local energy detection and cooperative information exchange. There are two kinds of models
to combine spectrum sensing and energy harvesting, i.e.,
time switching model and power splitting model. In the
time switching model, the SU operates as either a spectrumsensing receiver or an energy-harvesting receiver at any
given time, while in the power splitting model, the received
PU signal is split into two separate signal streams with different power levels, one used for spectrum sensing and the
other one used for energy harvesting, as shown in Fig. 5.
3.2.1 Time switching model
In the time switching model, the local sensing slot of the
frame structure is further divided into detection slot and harvesting slot. The SU firstly senses the PU in the detection
slot and if the presence of the PU is detected, the SU harvests the RF energy of the PU signal in the harvesting slot.
Then the harvested energy is used to compensate the energy
loss of data transmission in order to guarantee enough transmission power. Supposing that the time splitting factor is t ,
the lengths of harvesting slot and detection slot are (1t )
and t , respectively. Hence, the harvested energy is given
as follows
Eth = PH1 ps h2i (1 t )

(11)

Eth must be enough to compensate the energy consumption


for cooperative spectrum sensing, thus noting that the really
sensing time is t , we have Eth + E0 pe t + pc k
where E0 is the minimum initial energy supported by the
system battery, pe is the sensing power, pc is cooperative
power and k is cooperative time. Hence, the obtained residual energy for data transmission through the traditional and
proposed cooperative spectrum sensing is given by

E0 pe t pc k, traditional
(12)
Er =
E0 + Eth pe t pc k, proposed
Supposing that the frame length is T , the SU can transmit
data in the T transmission slot in the following two
cases:

Fig. 4 Energy-harvesting circuit

when the PU is really absent and the SU detects the


absence of the PU accurately in the probability 1 Qf ,

Mobile Netw Appl


Fig. 5 Energy harvesting-based
spectrum sensing

the SU can access the PU spectrum to transmit data


effectively.
when the PU is really present but the SU detects the
absence of the PU falsely in the probability 1 Qd , the
SU may also access the PU spectrum but cause harmful
interference to the PU.

Hence, the average spectrum access probability of the SU in


one frame is given as follows
(T k)
(PH0 (1 Qf ) + PH1 (1 Qd )) (13)
T
Our goal is to maximize the spectrum access probability
by jointly optimizing the sensing slot length , the number of cooperative SUs k and the time splitting factor t ,
subject to the constraints that the false alarm probability is
low enough, the detection probability is large enough and
the harvested energy can supply the cooperative spectrum
sensing. This optimization problem is given as follows

achieved only when Qd = . Noting that the sensing time


is t and substituting Qd = into Eqs. 10 and 13,the minimum Qf and the maximum PAcc are respectively given by



k
f
1
i2
(15)
Q = Q Q ()( + 1) + t fs
i=1

PAcc =

PAcc =



(T k) 
PH0 1 Q Q1 ()( + 1)
T



k
2
+ t fs
i
(16)
+ PH1 (1 )
i=1

From Eqs. 11 and 14d, we get that


t

ps + E0 pc k
(pe + ps )

(17)

where ps = PH1 ps h2i . Since from Eq. 15, Qf decreases


as t increases,PAcc may achieve the maximum only when
Eq. 17 acquires the equation. Hence, using ADO, we firstly
fix t with a constant 0 < t0 < 1 and solve the optimization problem about and k. With Qf we have

max PAcc

(14a)

s.t.Qf

(14b)

Qd

(14c)

Eth + E0 pe t + pc k

(14d)

Then the optimization problem (14) is rewritten as follows

0 T k

(14e)

maxPAcc =

1 k K, k Z

(14f)

0 t 1

(14g)

,k,t

where and are the upper limit of false alarm probability and the lower limit of detection probability, respectively,
K is the total number of SUs in CR network. From Eq. 7,
as Qf and Qd have the same monotonicity, PAcc increases
with the decrease of Qd , i.e., the maximum PACC can be

1 (t ) where 1 (t ) =

,k

(Q1 () Q1 ()( + 1))2



t fs ki=1 i2
(18)

(T k)
(P (H0 )(1Q(+(k) )+)
T
(19a)

s.t. 1 (t0 ) T k
1 k K, k Z

(19b)

(19c)


where = Q1 ()( + 1), (k) = t0 fs ki=1 i2 and
= P (H1 )(1 ). We use ADO to solve (19), i.e., we
optimize one of the two variables and k by fixing the other

Mobile Netw Appl

variable with an initial value, respectively, and the jointly


optimal solution can be obtained by repeating to optimize
these two variables and k until both of them are convergent
[13, 14].
Firstly, we initialize k = k0 , where k0 is any integer
within the limit [1, K], and select the k0 SUs with the largest
SNRs to perform cooperative spectrum sensing, then (k)
is a constant. Next, we will prove that when k = k0 is given,
Eq. 19 is a convex optimization problems, i.e., there exists
an optimal 0 [0, T k0 ] that makes PAcc ( ) achieve
the maximum. The first-order derivative of PAcc ( ) in is
given as follows

P (H0 )(1 Q( + (k0 ) )) +


PAcc ( ) =
T
(T k0 )P (H0 )(k0 )
+

2T 2


( + (k0 ))2
exp
(20)
2
where we further deduce that

lim < T1 (P (H0 )(1 Q())) < 0


T k0 

lim = O 1 = +

0 pc k
.
With the obtained and k, the optimal t = ps (p+E
e +ps )
Using ADO again, we obtain the jointly optimal solution
( , k , t ) to Eq. 19 using the Algorithm 2 through iteratively running Algorithm 1. Supposing that the estimation
accuracy is , the complexities of Algorithms 1 and 2 are
respectively given as O( 12 ) and O( 13 ).

(21)

where we have used the fact that Q(x) is a decreasing


function and upper bounded by 1. We also can get the
secondary-order derivative of PAcc in , which is related
with Qf as follows
2 PAcc ( ) =

2
T k0
P (H0 ) Qf ( )
P (H0 ) 2 Qf ( )
T
T
(22)

From Eq. 7, it is obvious that Qf ( ) < 0 and


Qf ( ) > 0. Thus we have 2 PAcc ( ) < 0. Therefore PAcc is convex in . Equation 21 means that PAcc ( )
increases when is small and decreases when approaches
T k0 , i.e., there deed exists a maximum point 0
[0, T k0 ]. 0 can be obtained by using the Newton gradient method. Hence, with the given k = k0 , the optimal
solution to Eq. 19 is given as follows
= max{0 , 1 (t0 )}

(23)

Then with given = , we optimize k to maximize


PAcc (k). Since k is an integer within [1, K], it is not computationally expensive to search through the optimal k from
1 to K. Hence, with the given = , the optimal solution
k to Eq. 19 is given as follows
k = arg max(PAcc (k)| = )

(24)

1kK

Using ADO, We obtain the jointly optimal solution


( , k ) to Eq. 19 using the Algorithm 1 through iteratively optimizing and k according to Eqs. 23 and 24.

3.2.2 Power splitting model


In the power splitting model, the whole local sensing slot is
used for energy detection. But the received PU signal is split
into harvesting signal stream and sensing signal stream. The
RF energy of the harvesting signal stream is converted to the
electric power to supply the cooperative spectrum sensing
which is then performed using the sensing signal stream.
Supposing that the power splitting factor is p , the powers
of the harvesting signal power and the sensing signal power
are (1 p )ps and p ps , respectively. Hence, the harvested
energy is given as follows
Eph = PH1 ps (1 p )h2i

(25)

Eph must be enough to supply the cooperative spectrum


sensing, thus we have Eph + E0 pe + pc k. Hence, similarly with Eq. 14, the joint optimization problem of power
splitting model is given as follows
max PAcc

,k,p

(26a)

Mobile Netw Appl

s.t. Qf

(26b)

Qd

(26c)

Eph + E0 pe + pc k

(26d)

0 T k

(26e)

1 k K, k Z

(26f)

0 p 1

(26g)

From Eqs. 25 and 26d, the maximum p is deduced as


follows
(ps pe ) + E0 pc k
(27)
p
ps
As the sensing signal power is p ps , the sensing SNR
is p . Hence, similarly with Eqs. 15 and 16, the minimum Qf and the maximum PAcc of splitting model are
respectively given as follows



k
f
1
i2
(28)
Q = Q Q ()(p + 1) + p fs
i=1

PAcc =



(T k) 
PH0 1 Q Q1 ()(p + 1)
T



k
2
+p fs
i
(29)
+ PH1 (1 )
i=1

Fig. 6 Residual energy of


different cooperative spectrum
sensing models

As PAcc also improves with the increase of p , the


maximum PAcc can be achieved only when p =
(ps pe ) +E0 pc k
. Hence, the optimization problem (26)
ps
can be solved similarly with Algorithms 1 and 2.
In a large scale network, a main node is selected, whose
distance to the PU is the shortest. Each node in the network senses the PU independently and then sends its local
sensing information to the main node. The main node will
make a final decision through using the proposed method to
combine these sensing information and broadcast the final
decision to all the other nodes [15, 16].

4 Simulations and discussions


In the simulations, K=50 SUs are randomly placed in a
100m100m square, the absence and presence probabilities
of the PU are P (H0 ) = 0.2 and P (H1 ) = 0.8, respectively, the channel gain h obeys the Rayleigh distribution
with the mean -10dB, the noise variance n2 =0.001mW,
the electromagnetism-to-electricity conversion efficiency is
=0.5, the sensing power is pe =0.2mW, the cooperative
power is pc =0.1mW, the average cooperative time of each
SU is =0.1ms, the frame length is T =10ms, the initial
energy is E0 =5mJ, and the convergency precision of the
algorithms is 103 .
Figure 6 compares the residual energy Er of the traditional and proposed cooperative spectrum sensing models
[6, 7], with different sensing time . It is seen that Er of the

3.6

3.4

3.2

2.8

traditional cooperative sensing,K=10


proposed cooperative sensing,K=10
traditional cooperative sensing,K=20
proposed cooperative sensing,K=20

E /mJ

2.6

2.4

2.2

1.8

/ms

10

Mobile Netw Appl


Fig. 7 Spectrum access
probability changed with
sensing time and user number
0.5
0.45
0.4
0.35

0.25

Acc

0.3

0.2
0.15
0.1
0.05
0
0

50
40

30

20

/ms

the sensing time and the number of cooperative users k.


It is seen that there deed exists an optimal set of and k
that maximizes PAcc . When =1ms and k=20, the maximum
PAcc =0.45. PAcc firstly improves and then decreases with
the increase of , because the sensing performance improves
while the transmission time decreases with , and therefore
there is a tradeoff between improving sensing performance

proposed model is larger than that of the traditional model.


In the proposed model, Er increases with the increase of
, however, in the traditional model, Er decreases with ,
because in the proposed model, the harvested RF energy of
the PU signal increases with and compensates the energy
wastage used for cooperative spectrum sensing. Figure 7
indicates the spectrum access probability PAcc changed with
Fig. 8 Spectrum access
probability changed with
sensing time with different
power of PU signal

10

0.45

p =50mW,Time switching
s

p =50mW,Power splitting

0.4

p =10mW,Time switching
s

ps=10mW,Power splitting

0.35

p =5mW,Time switching
s

ps=5mW,Power splitting

0.3

Acc

0.25

0.2

0.15

0.1

0.05

/ms

10

Mobile Netw Appl


Fig. 9 Spectrum access
probability changed with
sensing time with different
detection probability

Qd=0.9,Time switching
d

Q =0.9,Power splitting

0.6

Q =0.8,Time switching
Qd=0.8,Power splitting
d

Q =0.7,Time switching

0.5

Qd=0.7,Power splitting

Acc

0.4

0.3

0.2

0.1

and transmission performance. PAcc firstly improves and


then decreases with the increase of k, because the sensing performance improves while the cooperative overhead
increases with k, and thus there is also a tradeoff between
Fig. 10 Maximum spectrum
access probability versus
detection probability

/ms

10

improving sensing performance and decreasing cooperative


overhead.
Figure 8 shows the spectrum access probabilities PAcc
of the time switching model and power splitting model

0.65
d

Q =0.7,Time switching
Qd=0.7,Power splitting
0.6

Qd=0.8,Time switching
Qd=0.8,Power splitting
Qd=0.9,Time switching

0.55

Qd=0.9,Power splitting

Maxmimum P

Acc

0.5

0.45

0.4

0.35

0.3

0.25

0.2

0.8

1.2

1.4

Harvested energy /mJ

1.6

1.8

Mobile Netw Appl


Fig. 11 Maximum spectrum
access probability versus the
number of SUs

0.65

0.6

Maxmimum P

Acc

0.55

0.5

0.45

0.4

N=1,Time switching
N=1,Power splitting
N=5,Time switching
N=5,Power splitting
N=10,Time switching
N=10,Power splitting

0.35

0.3

0.25

0.8

1.2

1.4

1.6

1.8

Harvested energy /mJ

changed with the sensing time , with the power of the


PU signal ps ={50,10,5}mW. It is seen that compared to the
time switching model, PAcc of the power splitting model
is higher at a large ps but lower at a small ps , because
Fig. 12 Theoretic and practical
maximum spectrum access
probabilities

at small ps , the splitting sensing power is very little thus


yielding the lower sensing performance and decreasing the
spectrum access probability; however, at large ps , the splitting sensing power is large enough to guarantee the sensing

0.65

theoretic,fs=0.5kHz
practical,fs=0.5kHz
theoretic,fs=1kHz
practical,fs=1kHz
theoretic,fs=2kHz
practical,fs=2kHz
theoretic,fs=3kHz
practical,fs=3kHz

0.6

0.55

Acc

0.5

0.45

0.4

0.35

0.8

0.82

0.84

0.86

0.88

0.9

0.92

0.94

0.96

0.98

Mobile Netw Appl

performance, and since the harvesting and sensing of the


power splitting model are implemented simultaneously, the
single harvesting time is saved and thus the spectrum access
time can be improved. Figure 9 shows PAcc changed with
, with the detection probability Qd ={0.7,0.8,0.9}. It is seen
tha PAcc decreases with the increase of Qd , because the
false alarm probability Qf also increases with Qd , thus
decreasing the spectrum access of the SU. Hence, the maximum PAcc can be achieved only when Qd acquires its lower
bound.
Figure 10 shows the maximum PAcc of the time switching model and power splitting model changed with the
harvested energy, with Qd ={0.7,0.8,0.9}. It is seen that PAcc
decreases with the increase of the harvested energy, because
the sensing time or the sensing signal power reduces as
the harvested energy increases, hence, there is a tradeoff
between improving spectrum access and increasing harvested energy. Figure 11 shows the maximum PAcc changed
with the harvested energy, with the number of cooperative
SUs N={1,5,10}. It is seen that PAcc improves with the
increase of N, which indicates that the sensing performance
of cooperative spectrum sensing is outstanding and can be
improved through increasing the cooperative SUs.
Figure 12 compares the theoretical and practical maximum PAcc versus the sampling frequency
fs ={0.5,1,2,3}kHz, with different detection probability
Qd . It is seen that the practical maximum PAcc , which
is obtained by the proposed joint optimization algorithm
as shown in Algorithm 2, approaches the theoretical
maximum.

5 Conclusions
In this paper, we have proposed an energy harvestingbased weighed cooperative spectrum sensing to improve
the sensing performance and decrease the energy wastage,
which is implemented through two models: time switching
model and power splitting model. The time or the power is
divided into two parts, one for energy harvesting and the
other one for spectrum sensing, and the harvested energy
is used to supply the electric power consumed for cooperative spectrum sensing. Through jointly optimizing sensing
time, number of cooperative SUs and splitting factor, the
spectrum access probability of the SU is maximized while
guaranteeing sensing performance and transmission energy.
The simulation results have shown that (1) the residual
energy of the proposed cooperative spectrum sensing is
much larger than that of the traditional cooperative spectrum sensing; (2) compared to the time switching model, the
spectrum access probability of the power splitting model is
higher at a large power of the PU signal but lower at a small

power of the PU signal; (3) there is an optimal set of sensing time and number of cooperative SUs that maximizes
the spectrum access probability; (4) the practical maximum
obtained by the joint optimization algorithm accords with
the theoretical maximum.

Acknowledgments This work was supported by the National Natural Science Foundations of China under Grant Nos. 61471194 and
61301131, the Natural Science Foundation of Jiangsu Province under
Grant No. BK20140828, the Fundamental Research Funds for the
Central Universities under Grant No. NS2015088, the Chinese Postdoctoral Science Foundation under Grant No. 2015M580425, the
Scientific Research General Project of Liaoning Province Eduction
Commission under Grant No. L2014204 and the Scientific Research
Foundation for the Returned Overseas Chinese Scholars of State
Education Ministry.

References
1. Mitola J (2001) Cognitive radio for flexible mobile multimedia
communications. Mobile Netw Appl 6(5):435441
2. Hamdaoui B (2009) Adaptive spectrum assessment for opportunistic access in cognitive radio networks. IEEE Trans Wirel
Commun 8(2):922930
3. Ghasemi A, Sousa ES (2008) Spectrum sensing in cognitive radio
networks: requirements, challenges and design trade-offs. IEEE
Commun Mag 46(4):3239
4. Shen J, Liu S et al (2009) Robust energy detection in cognitive
radio. IET Commun 3(6):10161023
5. Duan DL, Yang LQ, Principe JC (2010) Cooperative diversity of
spectrum sensing for cognitive radio systems. IEEE Trans Signal
Process 58(6):32183227
6. Liu X, Tan XZ (2011) Optimization for weighed cooperative
spectrum sensing in cognitive radio network. Appl Comput Electromagn Soc J 26(10):800814
7. Zhao N, Yu FR, Sun H, Nallanathan A (2012) An energyefficient cooperative spectrum sensing scheme for cognitive radio
networks. In: Proceedings of the IEEE Global Communications
Conference. Anaheim, pp 36003604
8. Chau YA (2013) Energy-efficient cooperative spectrum sensing
with relay switching based on decision variables for cognitive
radio. In: Proceedings of the 16th international symposium on
wireless personal multimedia communications. Atlantic, pp 16
9. Valenta CR, Durgin GD (2014) Harvesting wireless power: survey of energy-harvester conversion efficiency in far-field, wireless power transfer systems. IEEE Microw Mag 15(4):108
120
10. Liu L, Zhang R, Chua KC (2012) Wireless information transfer
with opportunistic energy harvesting. IEEE Trans Wirel Commun
12(1):288300
11. Park S, Hong D (2013) Optimal spectrum access for energy harvesting cognitive radio networks. IEEE Trans Wirel Commun
12(12):61666179
12. Liu X (2014) A new sensing-throughput tradeoff scheme in cooperative multiband cognitive radio network. Int J Netw Manag
24(3):200217
13. Stephen B, Neal P et al (2011) Distributed optimization and statistical learning via the alternating direction method of multipliers.
Found Trends Mach Learn 3(1):1122

Mobile Netw Appl


14. Liu X, Jia M et al (2014) Optimal joint allocation of multislot
spectrum sensing and transfer power in multichannel cognitive
radio. J Sensors 2014:19
15. Huang X, Hu F, Wu J, Chen HH, Wang G, Jiang T
(2015) Intelligent cooperative spectrum sensing via hierarchical

dirichletprocess in cognitive radio networks. IEEE J Sel Areas


Commun 33(5):771787
16. Huang X, Wang G, Hu F (2013) Multitask spectrum sensing in
cognitive radio networks via spatiotemporal data mining. IEEE
Trans Veh Technol 62(2):809823

Anda mungkin juga menyukai